Conversation

1/3 Yesterday a test of an unreleased advanced order type encountered a bug which resulted in the order's prices being matched against the wrong side of the book. Some clients bought from the tester at $8000 and others sold at $12000 without clearing the intervening liquidity.
Image
65
124
205
1. Whose idea was to run tests on your production environment? 2. Which "clients" had a chance to buy BTC at $8000 and sell at $12000, while the market price was around $10250?
2
33
This Tweet was deleted by the Tweet author. Learn more
You’re arguing for going straight from test env to full public rollout without a production limited beta? You have to go to prod at some point. This particular feature had already seen thousands of tests over several months. This just reinforces our careful, staged approach.
9
25
Having "order's prices being matched against the wrong side of the book" is a bug of critical severity. Are you saying that "thousands of tests over several months" passed this test case on staging environment, but somehow failed in production?
2
9
No matter which "unreleased advanced order type" we're talking about here, your engine should simply not allow an order to be executed 20% above the lowest ask, or 20% below the highest bid. That's against the very fundamentals of how order books work.
1
6
Replying to and
This incident smells of a poor testing process combined with a poor matching engine. If at least ONE of the two was not poor, this would have been avoided. A good testing process would have discovered it on test or staging. A good engine would have rejected an incompatible order.
2
8